Reproduce “AgentEvolver: System-Wide Self-Evolution Through Task Execution” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wentao Zhang, Fuchao Yang, Yilei Zhao, Xinrun Wang, Bo An
AgentEvolver lets agents revise and reuse software capabilities without changing the foundation model. It versions components, tests candidates against preserved baselines, and records retention decisions and later use. On one rendered Godot frame, a brightness Tool’s median execution time fell 30.08%; SWE-bench Pro Public resolution was 82.08% versus 79.07%. The comparisons do not establish transfer, subsystem effects, or cost efficiency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.